https://github.com/julienpeloton/lafabrique

Generate simulated datasets for CMB experiments
https://github.com/julienpeloton/lafabrique

cmb simulated-datasets

Last synced: about 2 months ago
JSON representation

Generate simulated datasets for CMB experiments

• Host: GitHub
• URL: https://github.com/julienpeloton/lafabrique
• Owner: JulienPeloton
• License: gpl-3.0
• Created: 2017-04-20T13:05:02.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2017-06-21T11:16:55.000Z (over 7 years ago)
• Last Synced: 2024-11-13T10:15:19.076Z (about 2 months ago)
• Topics: cmb, simulated-datasets
• Language: Python
• Size: 1.7 MB
• Stars: 1
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
LaFabrique [![Build Status](https://travis-ci.org/JulienPeloton/LaFabrique.svg?branch=master)](https://travis-ci.org/JulienPeloton/LaFabrique)

==

![ScreenShot](https://github.com/JulienPeloton/LaFabrique/blob/master/additional_files/outputs.png)

## The package

This package contains scripts for generating simulated CMB datasets (current and future)

* Noise

    * Generate (inhomogeneous) noise maps

* Covariances

    * Generate covariance matrices (in the sense inverse noise weights)

* Foregrounds

    * Generate foregrounds maps (dust, synchrotron, freefree, spinningdust, ...) based on the PySM

* Frequency coverage

    * Multi-frequency

* Sky coverage

    * Scanning strategy simulator included

## Before starting

This code has the following dependencies (see the travis install section):

* numpy, pylab, scipy, etc (required)

* healpy >= 1.9.1 (required)

* h5py (required) + libhdf5 if using linux

* weave (optional if scipy <= v0.18 - required otherwise because no longer part of scipy v0.19)

* ephem (required - for scanning strategy)

* mpi4py (optional - for parallel computing)

* InsideMe (optional - for profiling of the code, see https://github.com/JulienPeloton/InsideMe)

* PySM (optional - foreground generation, see https://github.com/bthorne93/PySM_public)

## Installation

You can easily install the package using pip

```bash

pip install LaFabrique

```

Otherwise, you can clone the repo from the github repository and

use the setup.py for the installation. Just run:

```bash

python setup.py install

```

Make sure you have correct permissions (otherwise just add --user).

You can also directly use the code by updating manually your PYTHONPATH.

Just add in your bashrc:

```bash

LaFabriquePATH=/path/to/the/package

export PYTHONPATH=$PYTHONPATH:$LaFabriquePATH

```

## Usage

You can find ini files to have a quick usage:

```bash

python generate_dataset.py -h

python generate_dataset.py -setup_env setup_env.ini 

```

The code has a parallel architecture, which can be useful for large noise

simulations at high resolution (simulation of scan strategy and

foregrounds are done serially for the moment).

For example, just use (replace mpirun by your favourite one):

```bash

mpirun -n  python generate_dataset.py

    -setup_env setup_env.ini

    -setup_scanning setup_scanning.ini

    -setup_instrument setup_instrument.ini

```

Note that at NERSC, you have to replace the python binary by python-mpi which is

optimized for parallel computing.

Here is the breakdown of the code for 100 noise simulations for 8 frequency channels on 48 processors

at a resolution of nside = 2048, and a few percent of the sky. Notice that the

Cholesky factorisation is done only once per processor for each frequency band and

for all MC simulations (corresponds to 48[proc] * 8[freq] = 384 calls below).

![ScreenShot](https://github.com/JulienPeloton/LaFabrique/blob/master/additional_files/perf_100MC_nside2048.png)

## License

GNU License (see the LICENSE file for details) covers all files

in the LaFabrique repository unless stated otherwise.